Abstract My research in health economics has focused on how information and targeted consumer cost-sharing influences how patients choose providers and the financial savings of incentivizing patients to choose low-price providers. I have also examined the opposite side of the market, how patient use of information and targeted consumer incentives spurs provider price competition. These topics provided the framework for my research as a PhD student in Health Economics at the University of California, Berkeley and I continue to build on these topics while a policy researcher at the RAND Corporation. A natural next step for my career is to expand this line of research but in a more in-depth manner and using more advanced statistical methods. Performing mentored research in these areas will help me successfully make the transition from directed to independent research. The proposed study will help me to (1) contribute to a deeper understanding of patient health effects of an innovative insurance benefit design that is particularly relevant for the aging population; (2) continue to build capabilities working with large medical claims data sets and develop expertise in innovative statistical methods from different disciplines; (3) gain training in aging-related health-services research; (4) expand my exposure to the aging, health economics, and health services research communities; and (5) develop my abilities as an independent health services researcher and build the foundations to successfully compete for R01-level grants. In this project, I propose to examine whether reference pricing for colonoscopies and pharmaceuticals decreases adherence to recommended colorectal cancer screening and medication therapies among the near- elderly population. I will also examine the impact of reference pricing on patient health outcomes and the aging process. To do so, I intend to apply novel machine-learning statistical methods that have been recently developed in the computer science and statistics fields. As part of this proposal, I have built a formal training plan to develop expertise in these methods. This project will provide me with the flexibility and support to develop a long-term research agenda that focuses on using innovative statistical methods to evaluate the comprehensive effects of consumer cost- sharing programs. Although this study focuses on a single cost- sharing program, reference pricing, the skills I gain through this award will allow me to independently lead evaluations of future benefit designs. The application of machine-learning methods to the setting of reference pricing will provide a framework that I or other researchers can use to evaluate other insurance benefit designs or alternative patient populations. !